We thought that OpenAI’s Sora was light years ahead of the competition: Runway Gen-3 Alpha has just shown us otherwise

When OpenAI presented Sora in February of this year we were surprised. Although we had witnessed firsthand the evolution of image generation such as DALL·E, Midjourney or Stable Diffusion, we believed that high-quality video generation solutions would take a while to arrive.

We had reason to believe that the company led by Sam Altman was “light years” ahead of the competition. After all, just like ChatGPT, it had just showcased technology that was at a level never seen before. But in recent months the panorama has changed noticeably.

Runway Gen-3 Alpha and Sora’s competition

It turns out that OpenAI wasn’t that unattainable. In a matter of months, a variety of tools appeared on the scene intended to compete directly with Sora. Vidu and Kling AI emerged from China, while Dream Machine appeared from the United States and a few hours ago the new Gen-3 Alpha.

This last tool has been developed by the well-known New York firm Runway. Gen-3 Alpha arrives after Gen-1 and Gen-2 launched in 2023, but is billed as the first in a new series of models trained on a new multi-modal infrastructure. And the model boasts many new features.

Unlike previous Runway ML products, Gen-3 Alpha offers an improvement in three key aspects. On the one hand, according to what the company explains, you will have higher image quality. Visual representations may take on abstract or realistic tones, with an improved level of fidelity.

On the other hand, the company promises that its new model will also make a leap in terms of consistency. This point is interesting in any professional use where the aim is to achieve results that follow a line. It is of little use if the model is capable of generating great images, but different from each other.

Another highlight will be the movement. Here the creators of the models usually maintain a balance. The less movement there is, the less chance there is of unwanted patterns emerging in the video. With more movement, the risk increases, but Runway appears to be addressing it.

The model will be available in the coming days to Runway subscribers with advanced controls for the scene generation, such as Motion Brush, Advanced Camera Controls, Director Mode. To all this, OpenAI has not yet specified a release date for Sora. The company says it will arrive “this year.”

Images | runway

In Xataka | Sam Altman is considering turning OpenAI into a for-profit company, according to ‘The Information’. It would be his definitive turn